You have been asked to cut cloud spend by thirty percent without hurting reliability. What is your plan?

Get visibility first, allocating spend by service, environment and team, then work the list in order of effort and risk. Start with waste that has no reliability impact: idle resources, unattached volumes, over provisioned instances, non production environments running overnight, and log retention. Then commit to reserved or savings plans for the stable baseline. Only after that consider architectural changes, and take every proposal to the owning team with data.

I would not start cutting, I would start measuring, because without allocation by team and environment every conversation becomes an argument. Once I can attribute spend, I work in order of risk. The free tier of savings is pure waste: unattached volumes, old snapshots, idle load balancers, development environments running twenty four hours a day for a team that works eight, and log retention set to forever when compliance says ninety days. That alone has got me most of the way more than once. Next is right sizing based on actual utilization percentiles rather than what someone picked two years ago, done gradually with monitoring. Then commitments, so reserved capacity or savings plans against the baseline that clearly is not going anywhere, which is a discount for a promise rather than a change to anything. Only then do I look at architecture, things like data transfer patterns or moving cold data to cheaper storage, and those go to the owning team as a proposal with numbers. Last part is preventing regression: budgets, anomaly alerts and cost visible in the same dashboards as reliability.